Automatic Demand Response for load Scheduling and Management based on Customer Participation

Size: px
Start display at page:

Download "Automatic Demand Response for load Scheduling and Management based on Customer Participation"

Transcription

1 Auomac Demand Response for load Schedulng and Managemen based on Cusomer arcpaon Abhshek Goyal Deparmen of Elecrcal Engneerng Indan Insue of Technology Madras Chenna, Inda K. S. Swarup Deparmen of Elecrcal Engneerng Indan Insue of Technology Madras Chenna, Inda Absrac Demand Response s a key ngreden of emergng smar grds o keep he demand and supply mbalance n check. Need of Cusomer arcpaon s always fel bu s never fulflled o a sasfacory level. Ths paper proposes a Auomac demand Response model for cusomer parcpaon a he level of domesc consumers, o have an economcally beer consumpon paern accordng o her own preferences. The dea of auomac demand response (ADR) s proposed o nroduce he cusomer preferences and parcpaon n he load schedulng process. The workng of he model nvolves schedulng of loads n a day-ahead marke, gven he day ahead forecased load and generaor fuel cos funcon coeffcens. The workng of he proposed mehod s presened for a IEEE 30 bus sysem wh a normalzed day ahead load forecas daa Resuls ndcae he effecveness of he ADR approach for effecve load schedulng and managemen. Keywords: Auomac Demand Response (ADR), Cusomer arcpaon Model, Managemen, Dspach-able/ Non- Dspach-able, Shf Algorhm, Smar Grd., D g mn G, G, f c F ( g ) B NOTATIONS Lambda value a each hour Toal power demand Toal power generaed by un I Mn and Maxower lm for un Fuel cos coeffcens of un Gen fuel cos Hour Fuel cos funcon of un a g L u u f TL oa TLoal I. INTRODUCTION Dspachable load Sar me of dspachable load Run me of dspachable load fnsh me of dspachable load Dspachable load a hour Toal load a hour Toal dspachable load Toal load ALANCING of generaon and consumpon s a challengng ask n he effcen operaon of a power sysem. The uly company or generaor needs o provde enough generaon, ransmsson and dsrbuon capaces for meeng he peak load. As a resul, he power nework has a low load facor and s underulzed mos of he me [1]. The prce of generaon of elecrcy vares, bu he cusomers are charged a mean value averaged over a me perod. The load demand curve has peaks and valleys whch can be effecvely ulzed hrough load managemen o make fla profle, whch hs desrable. There s a need for a mechansm, whch can nsruc he generaon ules on how much o produce a each hour of he day ahead, whou oo many large flucuaons n demand. Accordngly, he amoun of operang reserve whch s opmally needed a ha demand can be decded. Ths makes he consumpon more ransparen and auomaed for he consumers, ermed as cusomer parcpaon. Ths decreases he overall operang cos of he sysem for boh consumers and generaors, whch s ermed as demand response. Demand response s an mporan aspec of smar grd research and has become a key ngreden of emergng smar grds o keep he demand and supply mbalance n check. Demand response s an mporan aspec of smar grd research. Implemenaon of smar grd s an mporan challenge faced n curren power scenaro. Demand Response echnologes allow ules o communcae o devces wh he cusomer premse. They nclude such hngs as load conrol devces, smar hermosas and home energy consoles. They are essenal o allow cusomers o reduce or shf her power use durng peak demand perods. Demand response soluons play a key role n several areas: prcng, emergency response, grd relably, nfrasrucure plannng, desgn and operaons. Whou an effecve DR here s no possbly of empowermen of he end-consumers no he power supply chan acves, hus s essenal o desgn a proper demand response scheme o effecvely shape a smar grd where boh he energy ules and end-consumers can benef. Ths paper addresses he mporan problem of demand response hrough cusomer parcpaon usng load managemen echnque, ermed as load shf or valley fllng algorhm. The concep of auomac demand response whch allows cusomer preferences s nroduced as an mporan conrbuon of he work. 1

2 II. DEMAND RESONSE IN SMART GRID The concep of smar grd sared wh he noon of Advanced Meerng Infrasrucure (AMI) o mprove Demand-Sde Managemen (DSM), energy effcency, and a self-healng elecrcal grd o mprove supply relably and respond o naural dsasers or malcous saboage [2, 3]. In bref, a smar grd s he use of sensors, communcaons, compuaonal ably and conrol n some form o enhance he overall funconaly of he elecrc power delvery sysem [4]. Tradonally power flow s almos undreconal from cenralzed supply sources o demand, and nformaon flow o operaonal ceners. In conras, n he emergng uly envronmen, boh power and nformaon flows are bdreconal. The emergng use of hermal sorage for peak shfng, he ancpaed growh and cos reducon of Solar hoo-volac (V) generaon a resdenal and muncpal levels, he ancpaed shf from convenonal fuel ransporaon o lug-n Elecrc Vehcles (EVs), he adven of low-cos smar sensors, and avalably of a wo-way secure communcaons nework across uly servce errory are ancpaed o sgnfcanly aler he naure of fuure power supply and power sysem operaons, as well as consumer behavor. III. AUTOMATIC DEMAND RESONSE (ADR) Demand Response (DR) s one mechansm whch manages cusomer consumpon of elecrcy n response o supply condons,.e. change n elecrcy usage by end users from her normal consumpon paern n response o changes n marke prces, or when sysem condons changes, or nework relably s jeopardzed [1]. The sysem condons nclude he prce of producon of elecrcy or he occurrence of any sae whch can lead he sysem no conngency. Demand response also ncludes acons aken by uly o respond o a shorage of supply for a shor duraon of me n fuure and response gven by he consumers o he above acons. Ths change n elecrcy consumpon can be rggered hrough manly wo mehods, Drec load conrol and Indrec load Conrol. Drec load conrol nvolves conrollng of he load drecly by he conrol cener. In case of hgh demand or low supply, he conrol cener on s own can swch on or off hs ype of loads accordng o he requremens. Indrec load conrol nvolves generang he nsrucons for he loads o be conrolled appropraely by he consumers hemselves. So, fnally all depends on he consumers hemselves wheher o follow hose nsrucons or no. So far here have been hree mehods of demand response o follow he above saed mechansms, whch are lsed as follows [2]. 1. Incenve Based DR rograms: Ths nvolves sendng ou of DR sgnals o he parcpang cusomers and hen provdng hem wh moneary ncenves when nsrucons are appropraely followed. I can be classfed under ndrec load conrol mehod. 2. Rae Based DR rograms: In hs program, cusomers are provded wh elecrcy prces so as o make her own nformed decson on when o consume more and when o keep a low load profle. Ths oo comes under ndrec load conrol. 3. Demand Reducon Bds: Ths caegory needs he cusomers o send DR bds o he aggregaor,.e. a wha prce hey agree o cural her some fxed amoun of consumpon. And hen accordngly he clearng akes place. Ths program s more of a drec conrol mehod. There exss ceran lmaons o he above mehods of conrollng he load, e.g. f drec conrol s appled o some load and he power supply o s los, hen wll no be able o execue, when he consumer prefers o be ON a ha parcular nsan. Smlarly f ndrec conrol s appled, hen he generang uly wll be expecng he load o be shu by he cusomer as soon as hey receve commands o do so. In a smlar way, wha f he consumer does no follow he nsrucons? The grd wll sll be runnng n a conngency sae. Ths gves rse o he need for a mechansm whch can ake cusomer preferences (gvng rse o cusomer parcpaon) respecng uly consrans and fnally schedules he load accordng o hese lms. Ths whole process of execung cusomer parcpaon and preferences, whle sasfyng uly consrans and execung load schedulng n an approprae fashon s referred o as Auomac Demand Response (ADR). Forecased Daa Generaor rcng Daa Generaon Consrans Cusomer references Demand Response Shf Algorhm ( Managemen) Levelng / Shfed aern STAGE I STAGE II STAGE III Fg.1. Basc workng model of Auomac Demand Response Fg. 1 shows he workng of he basc model of Auomac Demand Response (ADR). There are hree basc sages n he model of Auomac Demand Response. Sage 1 nvolves akng n all he consrans and preferences from he generang sde and he cusomer sde. In hs sage, he cusomer preferences are obaned from he load forecased daa, whle he generaor prcng daa and consrans are fed o he DR program n Sage 2. Sage 2 s he akng n of all he npus suppled a sage 1 and hen performng load schedulng process accordng o he sysem consrans. In hs sage, he load shfng algorhm of he DR usng he four npus from sage 1 provde a schedule and a shfed load paern o he consumer. Sage 3 fnally akes he shfed load paern from sage 2 and dspaches approprae sgnals o boh he generang sde and he consumer sde respecvely. 2

3 IV. CUSTOMER ARTICIATION MODEL Demand Managemen can be mos effcen only when here s a hgh level of nvolvemen from he source generang he demand, whch n our case s he cusomers or he consumers, also ermed as Cusomer arcpaon. Through hs, cusomers can become aware abou her consumpon level, can know he prce hey are payng and hence can ake her own wse decsons o consume elecrcy effcenly. They can n urn oban benefs n moneary erms, and due o he ransparency, he uly (sasfacon) level of cusomers can be rased o a very hgh level. As of now, AMI does no ncorporae he ably o program he power flowng owards dfferen loads, bu f smar meers are allowed o ac accordng o he sgnals comng from he Demand Response Conrol (DRC) Cener, so as o allow / block he power flow n dfferen load areas accordng o he nsrucons presen n he sgnal, hen dspachable loads can operae accordng o cusomer preferences. As an npu o he model, cusomer provdes hs me ahead preferences for dfferen loads, conneced o programmable nodes, gvng nformaon abou he run me lms, he acual run me, and he sze of he load conneced o each of he node. Ths can be done hrough a normal web poral based communcaon channel or hrough some AMI nerface. END USER / CONSUMER AMI Inerface Inpu I: Dspachable Daa Oupu I: Shfed aern Insrucons Demand Response Conrol Cener Inpu II: Generaor daa and day ahead load forecased daa Oupu II: Shfed load paern wh Economc Dspach Daa for Fnal Curve Generaon Company (GENCO) Transmsson Company (TRANSCO) Dsrbuon Company (DISCO) Fg. 2. Cusomer arcpaon n Auomac Demand Response (ADR). Fg. 2 shows proposed model of auomac demand response. The daa from he AMI s hen sen as an npu o DR conrol cenre wh/whou AMI communcaon channel, and hen processed wh all he daa colleced from all he cusomers, along wh he day ahead load forecased curve, so as o have a new load curve whch s much flaer and beer han he prevous one. DR conrol cenre wll hen generae he consumpon paern sgnal for he AMI, and also he generaon schedule for he generang uns as oupus. resenly, he workng of hs process was been esed and mplemened only for 24 hour day ahead scenaro akng 1 hour as he leas quanzaon of me, bu could be mplemened for lesser me perods as well wh approprae modfcaons. The processng a DR conrol cener nvolves, allocang he loads one by one o dfferen hours accordng o oal margnal cos of producon a each hour and he hours whn he run me lms so as o have he mnmum oal cos of producon requred for allocang ha parcular load. The ncluson of consumer parcpaon s proposed n he workng model, where cusomers segregae her loads no dspachable or non-dspachable ones [5-6]. Fg 3 shows he dspachable and non dspachable loads and her daa requremen o be used n he smulang ADR. Dspachable Non-Dspachable (a) Classfcaon Dspachable s 1. Sze of 2. Run Tme 3. Run Tme Lms (b) Dspachable daa requremen Fg. 3. Dspachable and Non-dspachable s Dspachable laods are he ones whch can be programmed or conrolled by he uly n cusomer preference lms, whle he non-dspachable are he ones where cusomers are he only conrollers and hence here s no nerference from he uly sde. The cusomers can now effecvely parcpae by gvng her preference for he dspachable load operaon. Ths daa s aken o demand response conrol cener hrough Advanced Meerng Infrasrucure (AMI), where hs daa, along wh oher daa of sysem consrans s operaed so as o have an opmum load schedule nsrucons o be sen back o AMI. V. MODELING AND IMLEMENTATION ASECTS The model of ADR descrbed above s llusraed here o allow he cusomer parcpae n elecrcy marke as a par of he Smar Grd Advanced Meerng Infrasrucure (AMI). In he modelng of he ADR, he followng daa was consdered o be avalable and provded: (1). Day ahead load forecased curve: Ideally only he oal load forecased curve s avalable. Ths model akes he concep of segregang each load no one of he wo caegores, he dspachable loads (rogrammable/ones wh flexble me of usage) and he nondspachable loads (non-programmable/ones whch are rgd n her me of usage). From he gven oal load forecased curve, calculaon of forecased curve for non-dspachable loads has been done assumng he rao of oal dspachable load o oal load, a each hour, consan.e. TL NL (1) Where, TL TL TL oal oal - Dspachable load a hour - Toal load a hour (2) 3

4 oal - Toal dspachable load for he day TL oal - Toal load for he day (2). Generaor cos funcon and ower lms: Generaor fuel cos funcon s assumed o be quadrac, whch wll have hree erms degree 2, degree 1, degree 0 (wh coeffcens,, respecvely). Generaon lms are specfed n erms of mum and mnmum power (, mn G G ) hey can delver once hey are chosen and swched ON afer he un commmen. These wo daa s are hen used o perform he economc dspach for he gven load level a each hour [7]. The cos funcon of a generaor s gven by: 2 mn F( ) f ( ), (3) g c g g Where, c G f represens he fuel cos. (3). Consumer reference daa for he dspachabale load: Each consumer eners he followng deals abou each of he dspachable load for o use n day ahead me lne: Value l, Up me u, Down me f and Run me r. g G g (a) Dspachable Daa Represenaon (b) Schedulng of Dspachanble loads usng Gan Char Fg. 5. Schedulng of Dspachanble loads under ADR Fgure 6 shows he flow char for he load shfng algorhm. (a) Sample Dspachable Daa (b) Gan Char for Dspachable Fg. 4. Dspachng loads usng Gan char. Fgure 4a shows he feaures of he dspachable load daa erms of sar me, run me and sop me. Fgure 4b shows he schedulng of dspachable loads usng Gan char. I can be observed from he varous possbles for schedulng he dspachable loads. The mplemenaon of he dspachng of loads s llusraed n fgure 5. Fg. 6. Flow char for Shfng Algorhm n ADR. Gven day ahead hourly load forecas, generaor prcng daa and oher sysem consrans, schedulng of he dspach able loads has been performed such ha our overall sysem cos s 4

5 mnmum, keepng he oher consrans of sysem secury whn lms.. The code was wren n MATLAB R2008a, and wndows 7 plaform was used for dong he same. VI. SIMULATION AND RESULTS An algorhm has been developed and successfully appled o boh a 6 bus, 3 generaors and IEEE 30 bus, 6 generaor sysems[appendx A1]. The s normalzed for a day ahead load forecas of a praccal sysem n Mllwood, USA (NYISO). Ths forecased daa was hen used o calculae he scalng facor wh base average as MW, and hen hs scalng facor was used for calculang forecased daa for our es sysems on a base average of MW. The oal amoun of energy lef as he dspachable one, s same as he oal amoun of energy our dspachable loads wll consume, as gven by he consumers n her preference. Toal no. of dspachable loads aken n hs case was Fg. 7. Shfng n Auomac Demand Response Fgure 7 shows he scheduled and shfed load based on he cusomer preferences wh ADR. Graphcal resuls llusrae ha he proposed mehod provdes a beer consumpon paern whou affecng he oal amoun of elecrcy consumed n a day. Imporan observaons, nference and conclusons are lsed below: Observaon: eak reducon has aken place afer separang loads no dspachabale and non-dspachable ones. The avalable 2250 loads have done he valley fllng. The fnal load curve (green) s more flaer han he nal curve. Inference: The dspachable loads have been shfed o off peak valley perods. Concluson: The new oal load curve (green) hus obaned, wh reduced peak level and flled up valleys, seems o have a lo beer profle hen he old curve. I has become lo flaer, whch was he man am, and here s no vsble hgh flucuaon whch can ncrease he probably of conngency. TABLE I DISATCHABLE / NON DISATCHABLE LOAD FORECASTED DATA Hour Mllwoo d Scalng Facor Forecased Non- Dspachable Fnal N o TABLE II SAMLE CUSTOMER REFERENCE AND ALLOTTED DATA Bus l Up me u (h) Down me f (h) Durao n r (h) Alloed sar hour Value (x10 W) Table II rovdes he daa for he sample cusomer preference for he 7 load buses from bus 14 o 20. The las column shows he resul of alloed sar hour for he cusomer o sar hs applances. TABLE III EFFECT OF LOAD SHIFTING ALGORITHM UNDER DEMAND RESONSE Hour IEEE 30 Bus Max. Mn Hour Average Sandard Devao n Faco r Orgnal Modfed Table III rovdes he effec of auomac remand response before and afer load shfng algorhm n erms of he comparson of he load curves correspondng o fgure 7. The analyss of he resuls are as follows: Observaon: Max has decreased, Mn load has ncreased, Average load sll he same, Sandard devaon has gone down by sgnfcan amoun, and load facor has ncreased by almos 1.7 % Shfng of loads has occurred and here are less flucuaons n he sysem. Concluson: Increase n load facor s a good sgn for effcency of power sysem and narrowng of gap beween mum and mnmum levels of load ndcaes he flaenng of he curve. 5

6 From he resuls, can be furher observed ha here has been mprovemen n he scheduled load n erms of reducon n he peak and fllng of valley, whle mananng he same average load wh apprecable reducon n he sandard devaon and load facor. I can also be observed ha he load curve has become much flaer, whch mples ha he auomac demand response has been effecve n load managemen echnques lke peak clppng, valley fllng and load shfng n makng he load curve relavely flaer. The mehod was mplemened on a sandard es sysem and can be easly exended for a praccal sysem. Ths work provdes cusomers o have more choces of paern of consumpon, whch leads o beer economc uly of consumer. I also provdes generaors wh a beer load curve so ha less no. of generang uns are nvolved (hgh sysem ulzaon) and hence he cosler ones can be shu off, also here be lesser probably of a conngency occurrng [7], [8]. Ths work can be a sarng sep for Independen Sysem Operaor (ISO)/Regonal ransmsson operaor (RTO) n lookng no he lowes level of dsrbuon nework (commercal and resdenal levels) for DR whch s nex challenge for smarer grds [3]. I can also be negraed wh oher DR programs o negrae varous oher aspecs of smar grd such as dsrbued generaon, large scale energy sorage and accordngly can be used o do perform dfferen power sysem schedulng asks as covered n [9-12]. CONCLUSIONS AND FUTURE WORK A mehod ha provdes an easy o use soluon for cusomers o parcpae n elecrcy cycle s proposed. shfng algorhm or valley cenrc algorhm, where mum loads were pu n he mnmum valley s more effcen han oher mehods. Ths work can be used for varous shapng objecves lke peak clppng, valley fllng shfng ec. Moreover, he proposed sysem could be furher mproved smply by usng more effcen code and subrounes and also can be esed for more no of loads. Ths work allows a sep forward o a Smarer Grd, where elecrcy can be reaed as a commody and can be used/sold/generaed lke any oher commodes Gen no AENDIX TABLE A1: GENERATOR FUEL COST FUNCTION DATA F f 2 ( g ) c( g g ) mn G G Fgure A1 shows he IEEE 30 Bus sysem consdered for he load shfng Algorhm. I s assumed ha load buses 14 o 17 are dspacable and can be programmed o faclae cusomer preferences and parcpaon. Table A1. Shows he correspondng generaor fuel cos daa used. Fg. A1. IEEE 30 Bus Sysem REFERENCES [1] The Smar Grd: An Inroducon, The US Deparmen of Energy, [2] Salman Mohaghegh, James Soups, Zhenyuan Wang, Zhao L, Hormoz Kazemzadeh, Demand Response Archecure, Inegraon no he Dsrbuon Managemen Sysem, Frs IEEE Inernaonal Conference Smar Grd Communcaons (SmarGrdComm), 2010, pp [3] Farrokh Rahm, Al Ipakche, Demand Response as a Marke Resource Under he Smar Grd aradgm, IEEE Transacons on Smar Grd, Vol. 1, No. 1, June 2010, pp [4] Clark W. Gellngs, The Smar Grd, Enablng energy effcency and Demand Response, CRC ress, The Far Mon ress, Inc, [5] hp:// - New York, Independen Sysem Operaors [6] eer B. Luh, Lauren D. Mchel, eer Fredland, Che Guan, Yung Wang, Forecasng and Demand Response, IEEE ower and Energy Socey General Meeng,, 2010, pp [7] Had Sada, ower Sysem Analyss, Taa McGraw Hll, [8] Frederck T. Morse, ower lan Engneerng, New Delh, Afflaed Eas Wes ress v. Ld, [9] L Zhang, Janguo Zhao, Xueshan Han, Ln Nu, Day-ahead Generaon Schedulng wh Demand Response, IEEE/ES Transmsson and DsrbuonConference & Exhbon: Asa and acfc Dalan, Chna, 2005, pp [10] edro Fara, Za A. Vale, Iude Ferrera, DemS - A Demand Response Smulaor n he conex of nensve use of Dsrbued Generaon, IEEE Inernaonal Conference Sysems Man and Cybernecs (SMC), 2010, pp [11] Masood arvana, Mahmud Fouh-Fruzabad, Demand Response Schedulng by Sochasc SCUC, IEEE Transacons on Smar Grd, Vol. 1, No. 1, June 2010, pp [12] J. Wang, S. Kennedy, J. Krley, A New Wholesale Bddng Mechansm for Enhanced Demand Response n Smar Grds, Innovave Smar Grd Technologes (ISGT), 2010, pp [13] hp://drrc.lbl.gov/ - Demand response research cener, Calforna energy commsson. [14] hp:// - Deparmen of energy, USA 6